Breath Analysis with Broadly Tunable Quantum Cascade Lasers

Jan 15, 2013 - Accurate isotope ratio determination of 12CO2 and 13CO2 in exhaled breath is of critical importance in the field of breath analysis, wh...
0 downloads 8 Views 1MB Size
Article pubs.acs.org/ac

Breath Analysis with Broadly Tunable Quantum Cascade Lasers Katharina Wörle,† Felicia Seichter,† Andreas Wilk,† Chris Armacost,‡ Tim Day,‡ Matthias Godejohann,§ Ulrich Wachter,∥ Josef Vogt,∥ Peter Radermacher,∥ and Boris Mizaikoff*,† †

Institute of Analytical and Bioanalytical Chemistry, University of Ulm, 89081 Ulm, Germany Daylight Solutions Inc., San Diego, California 92128, United States § MG Optical Solutions GmbH, 86922 Eresing, Germany ∥ Klinik für Anästhesiologie, Sektion Anästhesiologische Pathophysiologie und Verfahrensentwicklung, Universitätsklinikum Ulm, 89081 Ulm, Germany ‡

S Supporting Information *

ABSTRACT: With the availability of broadly tunable external cavity quantum cascade lasers (EC-QCLs), particularly bright mid-infrared (MIR; 3−20 μm) light sources are available offering high spectral brightness along with an analytically relevant spectral tuning range of >2 μm. Accurate isotope ratio determination of 12CO2 and 13CO2 in exhaled breath is of critical importance in the field of breath analysis, which may be addressed via measurements in the MIR spectral regime. Here, we combine for the first time an EC-QCL tunable across the 12 CO2/13CO2 spectral band with a miniaturized hollow waveguide gas cell for quantitatively determining the 12 CO2/13CO2 ratio within the exhaled breath of mice. Due to partially overlapping spectral features, these studies are augmented by appropriate multivariate data evaluation and calibration techniques based on partial least-squares regression along with optimized data preprocessing. Highly accurate determinations of the isotope ratio within breath samples collected from a mouse intensive care unit validated via hyphenated gas chromatography−mass spectrometry confirm the viability of IR-HWG-EC-QCL sensing techniques for isotope-selective exhaled breath analysis.

Q

conversion of glucose and glucose-6-phosphate shifts toward an increase of glucose-6-phosphate, which in turn inhibits glycogenolysis.18 As a result, the oxidation of glucose to CO2 within skeletal muscles is inhibited by switching from carbohydrate to noncarbohydrate utilization.19 Hence, noninvasively monitoring the glucose metabolism, for example, via exhaled breath biomarkers enables patient-specific medical care and early detection of sepsis and septic shock, respectively. The present study takes advantage of the fact that both CO2 isotopologues, 12CO2 and 13CO2, will be metabolized by the Krebs cycle. Hence, changes in the isotope ratio of exhaled carbon dioxide or metabolic intermediates such as alanine or lactate, for example, may directly be attributed to pathological metabolization processes. In order to detect a glucose metabolism dysfunction, 13C-enriched glucose was administered during the studies presented herein, which is partially metabolized, and finally exhaled as 13CO2. Consequently, changes of the so-called tracer-to-tracee ratio (TTR) (eq 1) in exhaled breath provide the first indications of a metabolic dysfunction.

uantum cascade lasers (QCLs) have seen a progressive development since their experimental introduction in 1994.1−4 With the introduction of a broadly tunable external cavity quantum cascade laser providing tuning ranges up to 525 cm−1 across the MIR spectral band,5 a wide variety of analytical applications may take advantage of such advanced light sources.5−12 Recently developed EC-QCLs tuning across the ν3-band of CO2 (2125−2525 cm−1) are of particular interest for application in medical diagnostics and exhaled breath analysis. Carbon dioxide isotopologues are diagnostically relevant parameters for monitoring several disease patterns. For example, observing the isotope ratio of the CO2 isotopologues 12 CO2 and 13CO2 in exhaled breath samples is used for diagnosing Helicobacter pylori infection13,14 or glucose metabolism dysfunction. The latter physiological condition is associated with septic shock, which is the most common cause of mortality in intensive care units in the United States.15 During the state of sepsis or septic shock, the energy demand of a patient strongly increases, which results in hyperglycaemia16 characterized, for example, by an enhanced consumption of peripheral glucose or de novo glucose synthesis.17 Metabolization of glucose is a sensitive cycle, which is immediately affected by increasing blood glucose levels during hyperglycaemia. Consequently, the glucose cycle regulating the © 2013 American Chemical Society

Received: October 22, 2012 Accepted: January 15, 2013 Published: January 15, 2013 2697

dx.doi.org/10.1021/ac3030703 | Anal. Chem. 2013, 85, 2697−2702

Analytical Chemistry

Article

⎛ c (labeled tracer) ⎞ TTR(%) = ⎜ ⎟100 ⎝ c (unlabeled tracee) ⎠

(1)

13 ⎞ ⎛ 13CO2 CO2 − 12 12 ⎟ ⎜ CO CO 2(sample) 2(standard) 13 δ C‰ = 1000⎜ ⎟ 13 CO2 ⎟ ⎜ 12 CO ⎠ ⎝ 2(standard)

(2)

trations, which is important for detecting small changes of δ13C‰. Combining QCLs with an external cavity and a movable grating nowadays offers a substantially enlarged tuning range compared to DFB-QCLs and facilitates evaluating broader absorption features in lieu of individual rotational−vibrational ones, thereby minimizing the effect of line-broadening processes during quantitative data evaluation. Hence, in the present study, a broadly tunable EC-QCL with a tuning range covering the ν3 band of 12CO2 and 13CO2 with a single device was applied. Aside from laser-based techniques, also Fourier transforminfrared spectroscopy (FT-IR) spectroscopic measurements are used for CO2 isotopic ratio determinations. Hofstetter et al.12 have shown a gas sensor prototype using an FT-IR spectrometer combined with a quantum cascade detector to exemplarily determine the 13CO2 concentration in ambient air and human breath at ppb levels based on a comparison with simulated spectra. The feasibility of determining the CO2 TTR by combining FT-IR spectroscopy with a hollow core waveguide33,34 (HWG) serving as both gas cell and light-propagating waveguide has recently been demonstrated by the research group of Mizaikoff.35−37 The developed HWG gas cell enables FT-IR and EC-QCL experiments in particularly small sample volumes (i.e., approximately 100 μL providing an optimized signal-tonoise ratio). While the utility of FT-IR-HWG systems is undoubted, usage for mouse intensive care units and future bedside patient monitoring requires a substantially smaller device footprint, as demonstrated for the EC-QCL-HWG sensor system herein. Due to the overlap within the ν3 band of 12CO2 and 13CO2 spectra, multivariate data evaluation based on partial leastsquares regression (PLS)38 using the SIMPLS algorithm39 was applied and the established calibration models were optimized for performance via appropriate data preprocessing routines. Consequently, the aim of the present study was the quantitative determination of the CO2 TTR in mouse breath samples using, for the first time, an EC-QCL-HWG gas sensor system along with validation of the results obtained via comparison to simultaneously recorded GC-MS data.

In medical studies, besides the TTR the so-called delta notation is common practice (eq 2). The δ value describes the isotopic abundance via a relative difference between the isotope ratio of a sample and a standard (i.e., most commonly the Vienna Pee Dee Belemnite20 in units of marked isotope per mill (δ13C‰). The determination of the CO2 TTR may be realized by several analytical methods. The most commonly applied techniques involve isotope-selective mass spectrometry frequently combined with gas chromatography (GC-MS)21−23 and isotope ratio mass spectrometry (IRMS).21,23 Both methods provide exquisite sensitivity and selectivity with IRMS representing the gold standard for low 13C enrichment levels in breath samples providing a linear response for samples up to a maximum enrichment of 924 δ13C‰.24 GC-MS appears more suitable for higher enrichment levels of >60000 δ13C‰, which is the range relevant for exhaled mouse breath samples resulting from the very low lung volume and body mass of mice. Furthermore, the determination of several volatile breath constituents is enabled by GC-MS. However, instrumental dimensions and sampling prerequisites do not facilitate continuous online measurements next to the patient or animal. Hence, analyzing the isotope ratio via spectroscopic or sensing techniques may supplement current exhaled breath analysis methods with online monitoring capabilities and, upon appropriate miniaturization, determination in minute sample volumes, as required for mouse breath analysis. Spectroscopic methods for the determination of CO2 and its isotopologues are frequently taking advantage of the isotopic shift associated with IR absorption features resulting from their difference in molecular mass. The most commonly applied technology in clinical use is nondispersive isotope selective infrared spectroscopy (NDIRS).25 NDIRS shows similar results as IRMS for enrichment levels of 1%. This phenomenon was already observed during previous studies using an FT-IR-HWG system and is subject to detailed ongoing research for quantifying and potentially correcting these effects.37 In Figure 3, a scores plot of the three latent variables illustrates their interrelation. 12CO2 and 13CO2 show an inverse behavior within the data space spanned by the three LVs. The

Table 1. EC-QCL Operating Parameters laser parameter operation mode scan mode laser current (mA) laser head temperature (°C) duty cycle (%) repetition frequency (kHz) pulse width (displayed) (ns) minimum wavenumber (cm−1) maximum wavenumber (cm−1) step size (cm−1)

internal pulse manual step 1453 15 5 100 500 2150 2450 0.5

Calibration samples of 13CO2 and 12CO2 were established via a gas mixing system (H. Wösthoff GmbH, Bochum, Germany) using a static mixing method based on induced turbulences between 50 mL syringes (B. Braun Melsungen AG, Melsungen, Germany). For evaluating the ν3 band of the CO2 isotopes 12CO2 (@ 2349 cm−1) and 13CO2 (@2283 cm−1), the spectral region of 2150−2450 cm−1 was scanned with the EC-QCL at a step width of 0.5 cm−1. For data evaluation, the target wavenumber was consistently applied (deviation from the measured 2699

dx.doi.org/10.1021/ac3030703 | Anal. Chem. 2013, 85, 2697−2702

Analytical Chemistry

Article

Figure 2. Expected vs predicted CO2 concentration, (left) 12CO2 and (right) 13CO2. Dots represent calibration samples and triangles test samples (quasi unknowns). Red line: linear fit and green line: y = x.

Figure 3. Scores plot illustrating the interrelation of the analyzed data within the three latent variables.

Figure 4. Radar plot of the scores of LV1 vs LV2. Black: mouse breath samples, green: 13CO2 calibration samples, blue: 12CO2, and red: 13 CO2-enriched samples. Figure 5. Radar plot of the scores of LV1 vs LV2. Black: mouse breath samples, green: 13CO2 calibration samples, blue: 12CO2, and red: 13 CO2 enriched samples.

13

C-enriched samples are located between the two branches of nonenriched samples and show a similar trend according to the 2700

dx.doi.org/10.1021/ac3030703 | Anal. Chem. 2013, 85, 2697−2702

Analytical Chemistry

Article

Figure 6. Summary of the obtained results for mouse breath samples independently analyzed on three different days: Gray, results EC-QCL-HWG data evaluated using the entire spectral range (QCL a); white, results EC-QCL-HWG data evaluated using spectral region selection (QCL b); and black, GC-MS data.

model are 0.961 for 13CO2 and 0.959 for 12CO2, respectively. Testing the predictability of the model offers a RMSEP of 0.256 for 12CO2 and of 0.051 for 13CO2. The RMSEC was determined at 0.242 for 12CO2 and 0.088 for 13CO2, respectively. The obtained calibration functions are comparable to the ones obtained using the entire spectral range (see the Supporting Information); likewise, the analyzed mouse breath samples distribute well within the set of calibration samples, as shown in Figure 5. Thus, obtained TTR of mouse breath samples along with the relative deviation to the GC-MS measurements are summarized in Figure 6. The obtained results reveal excellent agreement between the EC-QCL-HWG data and the GC-MS validation data. The mean relative deviation of the TTR determination is 3.4% (absolute mean deviation: 7.8%). Figure 6 gives a summary overview of all calculated TTR values based on EC-QCL-HWG data, along with the related GC-MS validation data. As there is no evident trend in over- or underestimating the TTR, systematic errors may be excluded. The results based on an evaluation of the entire spectral range (QCL a) reveal a slightly better compliance with the validation data compared to the results obtained after spectral region selection (QCL b).

grade of enrichment. The analyzed 18 mouse breath samples are likewise distributed in the regime of the 13C-enriched samples at the center of the model, yet, might be related to higher neat CO 2 concentrations rather than enriched calibration samples. Taking only the first two latent variables into account, a radar plot (figure 4) illustrates the analyzed mouse breath samples within a 95% confidence level. Outside the confidence level, 13 CO2 and 13CO2-enriched samples with a 13CO2 content of >1.5% and 12CO2 > 5% are located. Including these samples improves the resulting calibration model. The enriched samples are located parallel to the 13CO2 branch and are linked to the basic 12CO2 concentration. The resulting model was applied to determine the concentration of 13CO2 and 12CO2 within 18 collected mouse breath samples represented as the tracer-to-tracee ratio (TTR). For validation, samples of the same breath batch were analyzed via GC-MS; the obtained results are summarized in Figure 6. For these data and the developed data evaluation methodology, a mean relative deviation of −1.1% (absolute deviation: 5.8%) between the EC-QCL-HWG and GC-MS data was achieved, thereby corroborating excellent agreement between the ECQCL-HWG and GC-MS validation measurements. Evaluation after Spectral Region Selection. Due to the overlap of the ν3 bands of 12CO2 and 13CO2, an alternative approach for data evaluation after spectral region selection was investigated. For 13CO2, the spectral range of 2249−2270 cm−1 and for 12CO2, the spectral range of 2345−2360 cm−1 were selected. In general, slightly improved predictive capabilities were observed due to less perturbation by overlaying absorption bands of both isotopologues. Similar to the evaluation of the entire ν3 band, a set of optimized data preprocessing methods was applied in following sequence: (i) baseline correction using weighted least-squares, (ii) standard normal variate scaling, (iii) mean centering, and (iv) poisson scaling. In contrast to previously applied group scaling, poisson scaling scales the data by a factor S, which represents the square root of the mean of the data. Optimum results were achieved by reducing the considered concentration range to a maximum of 5% for 12CO2 and 2% for 13 CO2, respectively. The resulting calibration model was also based on 3 LVs and captures a spectral variance of 97.71% and a calibration variance of 95.96%. The obtained CODs for this



CONCLUSIONS The present study for the first time demonstrates the feasibility to determine the TTR of CO2 isotopologues (i.e., 12CO2 and 13 CO2) in small volumes (few hundred microliters) of mouse breath samples via a novel sensor system combining an external cavity quantum cascade laser with a miniaturized mid-infrared hollow waveguide gas cell. Validation by GC-MS analysis revealed excellent agreement with the obtained IR data. Appropriate multivariate data evaluation strategies (PLS) along with optimized data preprocessing steps successfully compensated for wavelength fluctuations inherent in the pulsed EC-QCL used in this study. While evaluation of the entire spectral window (approximately 400 cm−1) compared slightly better with the GC-MS validation data, spectral region selection offers substantially faster data acquisition rates, thereby rendering this newly developed IR sensing system ideally suited for future online monitoring of the TTR within mouse intensive care units. Finally, with the present study, EC-QCLHWG sensing systems have clearly demonstrated their 2701

dx.doi.org/10.1021/ac3030703 | Anal. Chem. 2013, 85, 2697−2702

Analytical Chemistry

Article

(17) Barth, E.; Albuszies, G.; Baumgart, K.; Matejovic, M.; Wachter, U.; Vogt, J.; Radermacher, P.; Calzia, E. Crit. Care Med. 2007, 35, S508−18. (18) Rossetti, L.; Giaccari, A.; Barzilai, N.; Howard, K.; Sebel, G.; Hu, M. J. Clin. Invest. 1993, 92, 1126−1134. (19) Vary, T. C.; Siegel, J. H.; Nakatani, T.; Sato, T.; Aoyama, H. JPEN, J. Parenter. Enteral Nutr. 1986, 10, 351−355. (20) Slater, C.; Preston, T.; Weaver, L. T. Rapid Commun. Mass Spectrom. 2001, 15, 1270−1273. (21) Barth, E.; Tugtekin, I.; Weidenbach, H.; Wachter, U.; Vogt, J.; Radermacher, P.; Adler, G.; Georgieff, M. Isot. Environ. Health Stud. 1998, 34, 209−213. (22) Amann, A.; Spanĕl, P.; Smith, D. Mini-Rev. Med. Chem. 2007, 7, 115−129. (23) Cao, W.; Duan, Y. Crit. Rev. Anal. Chem. 2007, 37, 3−13. (24) Wong, W. W.; Hachey, D. L.; Zhang, S.; Clarke, L. L. Rapid Commun. Mass Spectrom. 1995, 9, 1007−1011. (25) Haisch, M.; Hering, P.; Fuss, W.; Fabinski, W. Isot. Environ. Health Stud. 1994, 30, 247−251. (26) Vogt, J. a.; Wachter, U.; Mehring, J.; Radermacher, P.; Georgieff, M.; Fischer, H.; Holscher, U.; Moede, M.; Fabinski, W. J. Appl. Physiol. 2009, 107, 302−307. (27) Braden, B.; Gelbmann, C.; Dietrich, C. F.; Caspary, W. F.; Schölmerich, J.; Lock, G. Eur. J. Gastroenterol. Hepatol. 2001, 13, 807− 810. (28) Van Der Hulst, R. W.; Lamouliatte, H.; Megraud, F.; Pounder, R. E.; Stolte, M.; Vaira, D.; Williams, M.; Tytgat, G. N. Aliment. Pharmacol. Ther. 1999, 13, 1171−1177. (29) Murnick, D. E.; Peer, B. J. Science (Washington, DC, U.S.) 1994, 263, 945−947. (30) Faist, J.; Gmachl, C.; Capasso, F.; Sirtori, C.; Sivco, D. L.; Baillargeon, J. N.; Cho, A. Y. Appl. Phys. Lett. 1997, 70, 2670−2672. (31) Gmachl, C.; Baillargeon, J. N.; Capasso, F.; Sirtori, C.; Sivco, D. L.; Chu, S. N. G.; Cho, A. Y. IEEE Photonics Technol. Lett. 1997, 9, 1090−1092. (32) Rubin, T.; von Haimberger, T.; Helmke, A.; Heyne, K. J. Breath Res. 2011, 5, 027102. (33) Croitoru, N.; Dror, J.; Goldenberg, E. Fiber Integr. Opt. 1987, 6, 347−361. (34) Dahan, R.; Dror, J.; Croitoru, N. Mater. Res. Bull. 1992, 27, 761−766. (35) Wilk, A.; Seichter, F.; Kim, S.-S.; Tütüncü, E.; Mizaikoff, B.; Vogt, J. a; Wachter, U.; Radermacher, P. Anal. Bioanal. Chem. 2012, 402, 397−404. (36) Kim, S.-S.; Menegazzo, N.; Young, C.; Chan, J.; Carter, C.; Mizaikoff, B. Appl. Spectrosc. 2009, 63, 68A−86A. (37) Seichter, F.; Wilk, A.; Wörle, K.; Kim, S.-S.; Vogt, J. A.; Wachter, U.; Radermacher, P.; Mizaikoff, B. Anal. Bioanal. Chem. 2012, accepted for publication. (38) Wold, S. Chemom. Intell. Lab. Syst. 2001, 58, 109−130. (39) de Jong, S. Chemom. Intell. Lab. Syst. 1993, 18, 251−263.

potential as breath diagnostic devices, which may in the future be harnessed for direct online analysis of exhaled breath also in human intensive care analysis, therapy progression monitoring, or medication compliance testing.



ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: boris.mizaikoff@uni-ulm.de. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors acknowledge the support of the team operating the mouse intensive care unit (MICU) at the Klinik fü r Anästhesiologie, Sektion Anästhesiologische Pathophysiologie und Verfahrensentwicklung, Universitätsklinikum Ulm, and especially Sandra Weber for assistance during the mouse breath studies. The study protocol for mouse experiments was approved by the University Animal Care Committee and by the federal authorities for animal research of the Regierungspräsidium Tübingen, Baden-Württemberg, Germany.

■ ■

DEDICATION Dedicated to the memory of our friend and colleague Chris Armacost. REFERENCES

(1) Kazarinov, R. F.; Suris, R. A. Fiz. Tekh. Poluprovodn. (S-Peterburg) 1971, 5, 797−800. (2) Faist, J.; Capasso, F.; Sivco, D. L.; Sirtori, C.; Hutchinson, A. L.; Cho, A. Y. Science 1994, 264, 553−556. (3) Curl, R. F.; Capasso, F.; Gmachl, C.; Kosterev, A. a.; McManus, B.; Lewicki, R.; Pusharsky, M.; Wysocki, G.; Tittel, F. K. Chem. Phys. Lett. 2010, 487, 1−18. (4) Maulini, R.; Beck, M.; Faist, J.; Gini, E. Appl. Phys. Lett. 2004, 84, 1659−1661. (5) Caffey, D.; Radunsky, M. B.; Cook, V.; Weida, M.; Crivello, S.; Buerki, P. R.; Day, T. Proc. SPIE 2011, 7953, 79531K−79531K-11. (6) Young, C.; Kim, S.; Luzinova, Y.; Weida, M.; Arnone, D.; Takeuchi, E.; Day, T.; Mizaikoff, B. Sens. Actuators, B 2009, 140, 24− 28. (7) Kosterev, a. a.; Buerki, P. R.; Dong, L.; Reed, M.; Day, T.; Tittel, F. K. Appl. Phys. B: Laser Opt. 2010, 100, 173−180. (8) Brandstetter, M.; Genner, A.; Anic, K.; Lendl, B. Analyst (Cambridge, U.K.) 2010, 135, 3260−5. (9) Lu, F.; Belkin, M. a Opt. Express 2011, 19, 19942−7. (10) Farahi, R. H.; Passian, a; Tetard, L.; Thundat, T. J. Phys. D: Appl. Phys. 2012, 45, 125101. (11) Lewicki, R.; Doty, J. H.; Curl, R. F.; Tittel, F. K.; Wysocki, G. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 12587−92. (12) Hofstetter, D.; Di Francesco, J.; Hvozdara, L.; Herzig, H. P.; Beck, M. Appl. Phys. B: Laser Opt. 2011, 103, 967−970. (13) Goddard, A. F.; Logan, R. P. H. Aliment. Pharmacol. Ther. 1997, 11, 641−649. (14) Cao, W.; Duan, Y. Clin. Chem. (Washington, DC, U.S.) 2006, 52, 800−811. (15) Trimmel, H.; Luschin, U.; Köhrer, K.; Anzur, C.; Vevera, D.; Spittler, A. Shock 2012, 37, 140−4. (16) Mizock, B. A. Best Pract. Res., Clin. Endocrinol. Metab. 2001, 15, 533−551. 2702

dx.doi.org/10.1021/ac3030703 | Anal. Chem. 2013, 85, 2697−2702